For example, in a similar face image retrieval for searching for a face image similar to an object to be monitored, there is a technique of calculating a feature amount of a human face from an image captured by a camera or an image recorded by a recorder, comparing the feature amount with a feature amount of a face image of an object to be monitored, and extracting a human face close to the face image of the object to be monitored as a candidate image.
In the case of performing such a similar face image retrieval by, for example, an in-train monitor system, an image captured by a monitoring camera (train-side equipment) in a train is transmitted to ground-side equipment. In the ground-side equipment, a feature amount of a human face is calculated from a received image. The feature amount is compared with a feature amount of a face image of an object to be monitored, and a human face close to the human image of the object to be monitored is extracted as a candidate image.
As means of the train-side equipment to communication with the ground-side equipment, a self-owned radio, a cellular phone network, WiFi (Wireless Fidelity) provided in stations, and the like are considered. Generally, a self-owned radio is used and, in an emergency, a cellular phone network is used. However, since a self-owned radio has a narrow communication band, it is a communication network unsuitable for communication of a large amount of data.
Therefore, in the background art, a captured image is divided into an arbitrary size in accordance with the bandwidth of a communication network, and divided images are sequentially transmitted. In such a manner, although it takes time, an image of a large data size can be transmitted. However, in this manner, since it takes long time to transmit an image captured by a monitoring camera in a train to ground-side equipment, only a small amount of images captured by the monitoring camera can be transmitted to the ground-side equipment. As a result, the performance of a monitoring system cannot be improved.
The following patent literature 1 describes a technique of matching a face image captured by a monitoring camera against face images stored in a database.